The Infocomm Media Development Authority (IMDA) is a Singapore statutory board responsible for infocomm and media regulation, digital economy development, and AI governance. It operates under the Ministry of Digital Development and Information (MDDI) and is the author of Singapore's Model AI Governance Framework.
Overview
IMDA is a statutory board of the Government of Singapore. Its parent ministry is the Ministry of Digital Development and Information (MDDI). Its mandate covers infocomm and media regulation, development of the digital economy, and AI governance.
IMDA was constituted under the Info-communications Media Development Authority Act 2016, which establishes the Authority and defines the office of its Chief Executive (Info-communications Media Development Authority Act 2016 (Singapore)). The Act's long title describes it as establishing the Authority "and to make provision for competition and consumer protection in the media industry," and its substantive core is Part 7, conferring power to issue codes of practice alongside prohibitions on anti-competitive agreements and abuse of dominant position, consolidation review, directions, and dispute resolution, with Part 8 supplying investigatory and entry-under-warrant powers and offences for false or misleading information. The codes-of-practice power is the mechanism through which sector rules are made without primary legislation — the relevant feature for AI, since Singapore's measures issue as guidance and codes under this general converged-sector mandate rather than under a dedicated AI statute. The current version as at July 2026 incorporates amendments by Act 23 of 2025, in force June 29, 2026 (Info-communications Media Development Authority Act 2016 (Singapore)). IMDA describes its remit as developing and regulating the infocomm and media sectors, deepening regulatory capabilities for a converged infocomm-media sector, safeguarding consumer interests, and fostering pro-enterprise regulation (Source: https://www.imda.gov.sg/about-imda/who-we-are). Data protection in the private sector is administered through the Personal Data Protection Commission (PDPC), which operates under IMDA (Source: https://www.imda.gov.sg/about-imda/who-we-are).
Role in AI governance
IMDA leads Singapore's AI governance work, operating through the AI Verify Foundation, a non-profit, and through industry engagement. Its principal outputs include the Model AI Governance Framework, first published in 2019 with a second edition in 2020 as a non-prescriptive AI governance framework; the Model AI Governance Framework for Generative AI, published in May 2024 (Singapore Model AI Governance Framework for Generative AI (2024)); the AI Verify testing framework and toolkit, released in 2023; and the Gen AI Evaluation Sandbox.
IMDA describes its positioning as that of a "rule-taker and rule-bridge", designing Singapore's frameworks to interoperate with the OECD AI Principles, the G7 Hiroshima Code (G7 Hiroshima Code), the NIST AI RMF (NIST AI RMF), and the EU AI Act (EU AI Act) and GPAI Code of Practice (GPAI Code of Practice).
Model AI Governance Framework
Singapore released the first edition of the Model AI Governance Framework on 23 January 2019, providing guidance to private-sector organisations on ethical and governance issues in deploying AI. A second edition was released on 21 January 2020, adding considerations such as robustness and reproducibility and expanding the section on stakeholder interaction and communication. Both editions take a sector- and technology-agnostic approach and organise guidance around internal governance structures, the level of human involvement in AI-augmented decision-making, operations management, and stakeholder interaction (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence).
IMDA published companion materials alongside the Model Framework: an Implementation and Self-Assessment Guide for Organisations (ISAGO), developed with the World Economic Forum Centre for the Fourth Industrial Revolution and with contributions from more than 60 organisations; and a Compendium of Use Cases in two volumes documenting how organisations including DBS Bank, HSBC, Google, Microsoft, and Visa Asia Pacific aligned their practices with the framework. With the PDPC and the Lee Kuan Yew Centre for Innovative Cities, IMDA also published "A Guide to Job Redesign in the Age of AI" on 4 December 2020 (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence).
Model AI Governance Framework for Generative AI
The AI Verify Foundation and IMDA launched the Model AI Governance Framework for Generative AI on 30 May 2024 (Singapore Model AI Governance Framework for Generative AI (2024)), following a public consultation conducted between 16 January and 15 March 2024 (Source: https://www.morganlewis.com/pubs/2024/07/singapore-expands-its-ai-governance-approach-to-include-generative-ai). The framework expands the 2020 Model Framework, which covered traditional AI, to address risks the IMDA associates with generative AI, including hallucinations, copyright infringement, and value alignment (Source: https://www.morganlewis.com/pubs/2024/07/singapore-expands-its-ai-governance-approach-to-include-generative-ai).
The framework sets out nine dimensions to be considered in totality: accountability, data, trusted development and deployment, incident reporting, testing and assurance, security, content provenance, safety and alignment research and development, and AI for public good (Source: https://www.morganlewis.com/pubs/2024/07/singapore-expands-its-ai-governance-approach-to-include-generative-ai). The accountability dimension proposes allocating responsibility across model developers, application deployers, and cloud service providers based on each party's level of control, drawing on the cloud industry's shared-responsibility model and citing indemnity, insurance, and no-fault insurance as possible safety nets. The data dimension addresses data quality, the application of existing personal-data law to generative AI, privacy-enhancing technologies, and the use of copyright material in training datasets. The content-provenance dimension discusses digital watermarking and cryptographic provenance for labelling AI-generated content (Source: https://www.morganlewis.com/pubs/2024/07/singapore-expands-its-ai-governance-approach-to-include-generative-ai).
Model AI Governance Framework for Agentic AI
IMDA first published the Model AI Governance Framework for Agentic AI in January 2026, providing guidance to organisations deploying AI agents and stating that humans remain ultimately accountable. The framework is organised around four dimensions: assessing and bounding risks upfront by selecting appropriate use cases and limiting agents' powers; making humans meaningfully accountable through defined checkpoints requiring human approval; implementing technical controls across the agent lifecycle; and enabling end-user responsibility through transparency and training (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence). A May 2026 update (version 1.5) added real-world case studies and best practices on multi-agent systems, third-party agent risks, and automation bias (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence). IMDA also convened a working group of the Singapore legal community to produce a discussion paper examining how private-law frameworks apply to agentic AI, exploring fault-based and strict-liability models (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence).
AI Verify and the AI Verify Foundation
AI Verify is an AI governance testing framework and software toolkit developed by IMDA. The framework sets out 11 governance principles — including transparency, explainability, reproducibility, safety, security, robustness, fairness, data governance, accountability, human agency, and inclusive growth — and is described by IMDA as aligned with AI standards from the EU, the United States, and the OECD (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence). AI Verify was developed in consultation with companies including AWS, DBS, Google, Meta, Microsoft, Singapore Airlines, and Standard Chartered, released for international pilot in May 2022, and open-sourced in 2023, with more than 50 companies including Dell, Hitachi, and IBM participating (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence).
IMDA established the AI Verify Foundation, a not-for-profit body and wholly owned subsidiary, to develop open-source AI testing tools through the global open-source community (Source: https://aiverifyfoundation.sg/ai-verify-foundation/). The Foundation has nine premier members — AWS, Dell, Google, IBM, IMDA, Microsoft, Red Hat, Resaro, and Salesforce — and more than 180 general members (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence). The AI Verify Toolkit, hosted on GitHub, supports plugin-based extensions, including the Monetary Authority of Singapore's Veritas Toolkit (assessing models against FEAT principles) and the Competition and Consumer Commission of Singapore's AI Markets Toolkit (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence).
Testing tools and assurance initiatives
IMDA and the AI Verify Foundation maintain Project Moonshot, an open-source large language model evaluation toolkit that combines benchmarking and red-teaming and can be integrated into CI/CD pipelines to test for risks including hallucinations, undesirable content, data disclosure, and adversarial prompts (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence). In February 2025 the Foundation and IMDA launched the Global AI Assurance Pilot to codify emerging norms for technical testing of generative AI applications, alongside a Global AI Assurance Sandbox connecting builders and deployers of generative AI applications with specialist technical testers (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence). IMDA has also published a Starter Kit for Testing LLM-based Applications, providing pre-deployment testing guidance across five common risks: hallucination and inaccuracy, bias in decision-making, undesirable content, data leakage, and vulnerability to adversarial prompts (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence).
AI safety convenings
In partnership with Humane Intelligence, IMDA ran a multicultural and multilingual AI safety red-teaming exercise focused on Asia-Pacific in November and December 2024, bringing together more than 350 participants from nine countries to test four LLMs (Aya, Claude, Llama, and SEA-LION) for cultural bias in English and regional languages; a 2026 edition extended participation to 14 Asian countries and tested generative AI applications for data-leakage risks (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence). The 2025 Singapore Conference on AI brought together more than 100 AI scientists and produced "The Singapore Consensus on Global AI Safety Research Priorities", which organises AI safety research into three areas: development (creating trustworthy systems), assessment (evaluating risks), and control (monitoring and intervening after deployment) (Source: https://www.imda.gov.sg/about-imda/emerging-technologies-and-research/artificial-intelligence).
The constituting statute is tracked at Info-communications Media Development Authority Act 2016 (Singapore).